On 18/03/2014 4:26 a.m., Kinkie wrote: > Hi all, > with the advent of 64-bit systems, which arguably are now most of our > install base, mallinfo(3) has started to show its age. According to comment > threads in various mailing lists, that library call is old, not really > portable, and will not be updated to 64-bit cleanliness even on Linux/glibc > systems. Our user mailing list has more examples of it causing confusion > due to 32/64 bit wrapping than helping people out. > This patch removes all references to mallinfo from the code. > > If all the patches I just posted are +1-ed for inclusion, they can all be > merged in one go from the lp:~kinkie/squid/sbrk-removal feature-branch. > > Thanks >
I'm favour of this and have been for some time. However the patch does not go far eonough to do what this submission claims. 1) do_mallinfo is removed in tools.cc but the globals.h and main.cc parts of it are not removed. 2) The high_memory_warning directive now depends uniquely on HAVE_MSTATS && HAVE_GNUMALLOC_H. * Please add an cf.data.pre IFDEF check on those macros being defined * Please update the directive documentation to omit mallinfo and define the documentation. * perhapse a TODO entry about linking that directive to a mempools capacity limit would be good idea. 3) was --enable-snmp used at the time? IIRC this should be interacting with some of the memory OID stats but I don't seen any OID calculatiosn or docs being adjusted. Amos
